Can you expand on the use case you have for this benchmark? Designing around a specific values could lead to some difficulties later.
Some examples dimensions that could introduce latency to such a measurement:
- Network conditions between the device and cloud.
- The distributed nature of the cloud.
- Accuracy of HW timing.
- Extreme outliers in the dataset. (Timeouts for example).
One possible approach to timing this connection, is to send an MQTT publish of QoS 1, and recording the amount of ticks between sending the publish and receiving the ACK. The ticks can then be converted to a unit such as milliseconds.
Hope this is helpful.